Re: [PATCH] sys_remap_file_pages: fix ->vm_file accounting

From: Oleg Nesterov
Date: Mon Feb 11 2008 - 05:16:53 EST


Sorry for delay,

On 02/06, Hugh Dickins wrote:
>
> On Sun, 3 Feb 2008, Oleg Nesterov wrote:
> > Off-topic question to all. sys_remap_file_pages() doesn't work with
> > shared readonly mappings, why?
>
> Slight correction: it works with shared readonly mappings, doesn't
> it, so long as the mmap'ed file was opened for reading and writing?

Yes sure. I meant, if the file was opened without FMODE_WRITE, then
mmap(PROT_READ, MAP_SHARED) doesn't actually set VM_SHARED, it only
sets VM_MAYSHARE.

(this looks understandable, but means that !VM_SHARED doesn't necessary
imply the possible cow).

> > IOW, why it checks VM_SHARED but not VM_MAYSHARE?
>
> My guess has always been that it was just a misunderstanding of how
> those VM_ flags end up working: assume so unless Ingo corrects me.
>
> By the time I realized that oddity, we'd been driven into several
> tiresome corners by the very existence of sys_remap_file_pages.
> So whereas my usual instinct would have been to relax the restriction
> and generalize, in its case I wanted to hold on to every restriction
> we had.

OK, thanks a lot. I was afraid I missed some "obvious" reason why we
can't do this.

Oleg.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/